I want to remove watermark from an image. I have found an api that performs exceptionally well for my usecase(https://clipdrop.co/apis/docs/inpainting). This api accepts image and a mask of same size as multipart/form data. The mask image has to be a transparent image of same resolution as the image, with a white patch to indicate the object to be removed. I want to make an api extending this where I can upload multiples images and one logo that needs to be removed, instead of the mask. This allows a bulk removal of watermark. The challenging part is addressing different resolution size of image and different size/locations of the logo.
I have tried using template matching with openCV to create a mask for each image and pass it to the api, but it doesn't seem to work. It works only when the supplied logo is a clip of the original image. I can't use a png of my logo and expect it to work. Is there any workaround for this task? Or a completely different tool that's already been made? Any help/suggestions would be appreciated.